Case Study

Lenfest AI Fellow spotlight: Dana Chiueh

ProPublica

October 19, 2026

This profile is part of the Lenfest AI Collaborative and Fellowship Program’s 2026 Impact Report.

As an AI engineering fellow at ProPublica, Dana Chiueh is exploring how artificial intelligence can help investigative newsrooms better manage and act on information from their communities. We spoke with Dana about the lessons she’s learned so far, what excites her most about AI, and where she believes the industry is headed next.

“I think the implementation of LLMs should only reinforce the argument for having more journalists on the ground,” she said, pointing towards a misconception that LLMs can take over journalists’ work.

About her role and the problem she is trying to solve

Dana’s background bridges data reporting and computer science, a combination that naturally led her into AI-focused roles in newsrooms. After studying at Oxford and Stanford, she is currently an AI engineering fellow at ProPublica and previously held a similar role at The Minnesota Star-Tribune as part of the Lenfest AI Collaborative and Fellowship.

Her core mandate at ProPublica is to explore how machine learning, AI, and other emerging technologies can be applied to engagement reporting. In practice, that means tackling a hard, very real newsroom problem: How to manage and honor the wealth of information that pours in from the public.

Newsrooms, particularly investigative ones, receive a large volume of tips, messages, and source outreach. Some of these are sensitive and need to be handled with privacy in mind. Sorting, prioritizing, and maintaining relationships with these sources at scale is difficult:

It’s hard to process and triage a high volume of incoming tips.
It’s hard to make editorial decisions about which tips to pursue when information is incomplete.
It’s hard to sustain relationships over time with so many people writing in.

“A big part of my mandate is just to figure out: Can artificial intelligence help us with this set of problems?” Dana said.

She is also working more broadly on internal reporter tools, exploring ways to thoughtfully introduce generative AI and machine learning into daily workflows inside the newsroom.

What has surprised her most during the fellowship

When Dana looks back over the past few years, what stands out is how quickly a community around AI and journalism has sprung up.

During her master’s at Stanford, when she was already exploring the intersection of AI and journalism and building newsroom software, that area felt “pretty nascent.” There were only a handful of people in the field, and use cases for tools like ChatGPT were still being worked out.

Since then, she has seen the emergence of:

Pop-up communities of practice
Events and gatherings focused on AI in journalism
A much larger pool of people working on similar problems across newsrooms

She notes that in the second year of the fellowship, it feels like there is “a lot more wind in our sails,” with more momentum, collaboration, and funding supporting this work.

At the same time, that growth brings a new challenge: bridging the divide between editorial and product/technical practitioners working with AI. Dana is especially interested in how to foster stronger technical collaboration across newsrooms, so that people aren’t solving the same problems in isolation.

What’s one thing the industry gets wrong about AI

From Dana’s perspective, there’s a misunderstanding about what AI is actually good at and, importantly, what it is not a substitute for.

She sees a clear distinction between:

AI’s strength in “repeating the past,” such as summarizing existing information, surfacing historical context, and accelerating research.
Journalism’s core mission of finding new facts and being deeply embedded in communities.

What makes journalism unique is human reporting, newsgathering, knocking on doors, and discovering what’s happening right now. AI can accelerate research and context-building, but it cannot replace human relationships and on-the-ground reporting, she stressed.

One lesson from her work that other newsrooms could apply tomorrow

A key lesson Dana has learned is the importance of involving stakeholders early and deeply in AI product development.

AI products, in her experience, are inherently collaborative. They often need:
Editorial partners to help define the problem precisely.
Journalists to label data, classify examples, or manually perform tasks once so that their judgment can be modeled.
Ongoing feedback to refine and evaluate tools.

She suggests that newsrooms:

Use AI to quickly build very lo-fi or wireframe prototypes.
Put these early versions in front of stakeholders to spark imagination and concrete feedback.
Make it clear from the outset that AI tools will require significant editorial involvement, not just a handoff to product and engineering for a few months.

In other words, treat AI tools as co-created editorial products, not standalone technical projects.

What she is paying attention to right now

One emerging area that fascinates Dana is adaptive interviewing.

She’s been following research on using adaptive interviewing for areas like UX research, user testing, market research, and polling, and has also seen it applied in the context of journalism.

It involves question flows that change based on a person’s previous answers, making forms or bots more personalized, dynamic, and responsive to user preferences.

She is intrigued by what this might mean for reporting and engagement at scale, while also recognizing the existential questions it raises about what it means to interview and connect with someone in journalism.

“When I think about how we manage relationships at scale, it’s certainly a compelling idea that piques my curiosity,” she said.

What she is most optimistic about

Dana is optimistic about the future of shared, collaborative AI infrastructure for newsrooms.

She sees growing recognition that:

AI can significantly help with newsroom workflows.
There is a need for technology that outlives specific time-limited engagements.
Third-party organizations, such as foundations or industry associations, could help maintain and distribute shared tools across multiple newsrooms.

She believes there is a real opportunity for industry-wide collaboration, where tools developed in one newsroom are sustained and adapted for others, moving beyond one-off open-sourcing toward durable, shared systems that benefit the whole ecosystem.